Globally, hypertension is the leading preventable risk factor for cardiovascular disease and all-cause mortality (Mills et al, 2020). Hypertension accounts for over 10.8 million premature deaths worldwide (Vaduganathan et al, 2022). In 2021, the all-cause disability-adjusted life years due to hypertension were 2770 per 100 000 (Vaduganathan et al, 2022). According to the World Health Organization (WHO), it is estimated that 1.3 billion adults between 30 and 79 years of age have hypertension. Furthermore, approximately 46% of hypertensive patients are unaware that they have the condition and less than half (42%) are diagnosed and treated (WHO, 2023). Among those with hypertension, more than half have additional cardiovascular risk factors (Unger et al, 2020).
Community nurses often identify the issue (Clark et al, 2000). Previous studies have shown that community nurses are well placed to provide effective education and support to improve patient adherence to remote blood pressure monitoring (Artinian et al, 2007; Cooper and Zimmerman, 2017; Simonetti et al, 2021).
Hypertension remains a significant risk factor for heart failure (Oh and Cho, 2020), atrial fibrillation (Verdecchia et al, 2018), stroke, myocardial infarction, chronic kidney diseases and dementia (Lee et al, 2022). Globally, heart failure affects approximately 64 million people, with more than 15 million cases in Europe (Conrad et al, 2018). Substantial economic burden is associated with hypertension; direct healthcare costs, productivity losses, and the financial toll on individuals and healthcare systems (Jeemon et al, 2021). This makes hypertension a growing public health challenge that requires innovative solutions. Several factors behind this increasing challenge have been identified as: an ageing population, increase in life expectancy, population growth and sedentary lifestyles (Murray et al, 2020).
Innovative strategies aimed at promoting early screening, detection and effective treatment of hypertension remain crucial in efforts to attain the WHO's goal of a 33% reduction in hypertension cases by 2030 (WHO, 2023). Nurse-led community-oriented screening has proven effective in the early detection of hypertension and in encouraging patients to pursue further medical attention (Lucky et al, 2011; Adler et al, 2019). In community healthcare settings, the use of traditional cuff-based devices for measuring blood pressure remains the predominant method of diagnosing hypertension and monitoring patients over time (Unger et al, 2020; Mancia et al, 2023; National Institute for Health and Care Excellence (NICE), 2023). However, cuff-based blood pressure devices are not suitable for continuous real-time measurement and have several other limitations (Gircys et al, 2019; Lee et al, 2019). In a move to effectively enhance early diagnosis, monitoring and treatment, cuffless devices have been developed using the latest technology to monitor blood pressure in real time (Goldberg and Levy, 2016).
The development and use of cuffless devices that allow for continuous real-time measurement and remote monitoring may promote wider uptake, enhance patient autonomy and give community nurses more information about their patient's blood pressure profile, potentially leading to improved blood pressure control and better long-term clinical outcomes (Bard et al, 2019; Blood et al, 2023). The validity and effectiveness of these cuffless devices in a clinical setting has not been examined exhaustively, as concluded by guidance from the European Society of Hypertension (Mancia et al, 2023).
Aim of commentary
This commentary aims to critically appraise the methods used within the review by Islam et al (2022) and expand on the findings in the context of community nursing and clinical research.
Methods of the review by Islam et al (2022)
This systematic review undertook a comprehensive search of published and unpublished studies from a range of electronic databases, using the controlled vocabulary of Boolean operators. No restriction on publication type or language was applied to the search. Further searches of electronic databases were supplemented by hand-searching the reference lists of included studies for potentially missed eligible studies and grey literature. Two reviewers (SMSI and NS) independently screened the titles and abstracts of the search results from relevant databases.
A predefined inclusion criteria was applied; studies had to describe the validation of wearable cuffless blood pressure devices against a reference device in humans. The review excluded studies that: did not provide a reference device for validity assessment; reported devices with invasive sensing components or a pneumatic cuff; evaluated devices with animal simulation models; provided only algorithms; lacked full text. Additionally, studies that provided only methodological principles or insufficient details to establish eligibility were excluded. Abstract and title screening were undertaken independently by two reviewers (SMSI and NS). Full paper screening and data extraction were reportedly carried out in duplicate; however, it remains unclear whether this process was conducted independently.
Quality assessment was undertaken independently by two reviewers (RD and SMSI), with arbitration by a third reviewer (CK). The Quality Assessment of Diagnostic Accuracy Studies version 2 (QUADAS-2) tool was used to evaluate any risk of bias. Studies reporting the mean absolute difference between test and reference of 5 mmHg for both systolic blood pressure and diastolic blood pressure were meta-synthesised using a random effects model.
Results
After removing duplicates, 418 records were identified and 16 were included in the systematic review after screening. The majority of studies used photoplethysmography (n=13) to assess blood pressure; the remaining three studies used digital auscultation, magneto-plethysmography and seismocardiography. Devices were used at a range of anatomical locations. Gold standard index test was not used for comparison, although all index tests were classified as approved devices within the review. The devices for commercial processing varied in cost, ranging from $50 to $3000 USD. Eight of the studies successfully showed a mean bias of under 5 mmHg when comparing systolic and diastolic blood pressure to the reference device. Only six studies were included in the main meta-analysis.
Comparison of different cuffless wearable blood pressure devices and various reference measuring devices, showed no evidence of difference in systolic (MD 3.42 mmHg; 95% CI -2.17 to 9.01; n= 6 studies) or diastolic (MD 1.16 mmHg; 95% CI -1.26 to 3.42; n=6 studies) blood pressure readings. However, there was a statistically significant and substantial difference between study heterogeneity for both systolic (I2=95%) and diastolic estimate (I2=87%). The prediction interval for both systolic (-11.92 to 18.76 mmHg) and diastolic blood pressure (-4.82 to 7.15 mmHg) suggests that error rates greater than 5 mmHg cannot be ruled out.
This review also carried out two sensitivity analyses on an additional five datasets, with no clear explanation of why these were excluded in the main meta-analysis. Moreover, two datasets were used from a single sample of the same participants, breaking the rule of statistical non-independence. In the take-one-away analysis of a single study, which was expected to show a large bias, the pooled mean bias of 11 datasets was 3.16 mmHg, standard deviation (SD) 4.13 for systolic and 1.22 mmHg, SD 2.25 for diastolic blood pressure. In the take-two-study-away analysis, which was expected to show a large bias, the pooled mean bias of 10 datasets was 2.54 mmHg, SD 4.21 for systolic and 0.93 mmHg, SD 2.22 for diastolic blood pressure.
Similarly, in the subgroup analysis of wearable device type sensor, nine additional datasets were used, which were not in the main analysis. The mean bias for studies that assessed wearables with photoplethysmography sensors was 12.09 mmHg, SD 14.30 for systolic and 3.27 mmHg, SD 2.25 for diastolic blood pressure (n=5 studies). The mean difference for studies that used photoplethysmography and electrocardiogram was 2.18 mmHg, SD 1.01 for systolic and 0.40 mmHg, SD 1.56 for diastolic blood pressure (n=4 studies).
Commentary
Using the Joanna Briggs Institute's critical appraisal tool for systematic reviews and research syntheses, Islam et al (2022) was appraised. Out of the 11 criteria, three were not achieved (Aromataris et al, 2015). The methods for full-paper screening and data extraction lacked clarity. While it seemed that these tasks were conducted in duplicate, the absence of an explicit description of a moderating process raised uncertainties. The absence of duplication and independence in these processes could potentially result in relevant studies being overlooked and errors during data extraction (Buscemi et al, 2006; Waffenschmidt et al, 2019).
The second issue was the lack of clarity regarding the methods used for synthesising the findings of the included studies. For example, the article indicated that only six studies were included in the main meta-analysis, yet the sensitivity analysis had 11 studies. The authors conducted the main analysis using statistical product and service solutions (SPSS) software and MATrix LABoratory (MATLAB) platform in the subgroup and sensitivity analysis, without explanation of why two different software types were used. The lack of indication regarding the individual weightings of studies within the subgroup and sensitivity analyses makes it challenging to discern the specific modelling or weighting assigned to each study. This diminishes the repeatability of the methods employed for this analysis. The rule of statistical non-independence was breached, with two datasets coming from a single study (Zheng et al, 2014). Within this study, the same 10 participants were assessed twice using two different methods. This could lead to a possible type I error (false positives) (Nakagawa et al, 2017). It is important to note that this only occurred within the sensitivity analyses, not included in the six studies used in the main analysis.
The final criterion that was not met was the absence of an assessment of publication bias. While this might be less applicable in the main analysis comprising only six studies, the inclusion of an additional 11 studies identified in the sensitivity analysis could have made such an assessment more relevant. Consequently, this raises the potential issue of publication bias within this literature.
There were also concerns about the quality of the studies included in the review. Only six reported the use of a standardised international blood pressure validation protocol. The absence of a standardised protocol for assessment complicates repeatability and could be a potential cause of heterogeneity. European Society of Hypertension recommends that cuffless blood pressure devices should undergo six validation tests, depending on the type of device and function (Stergiou et al, 2023). Based on the methodological and study-related issues, these findings should be viewed with some caution. However, with the current evidence base from this review, the recommendations given below for future research are warranted. The goal is to enhance the importance of technology in healthcare and its ability to adapt to changing needs.
These findings use a mean error encompassing both systolic and diastolic measurements, which complicates comparisons with the current validation guideline procedures. Previous guidelines from the Association for the Advancement of Medical Instrumentation (AAMI)/International Organization for Standardisation (ISO) and the European Society of Hypertension—International Protocol, advocate for recommendations grounded in the likelihood of an error, with a mean bias of ≤10 mmHg occurring at least 85% of the time (O'Brien et al, 2010; Stergiou et al, 2023). As noted in the review, eight studies reported a mean bias for both systolic and diastolic measurements of less than 5 mmHg, but there was no indication of the total number of studies or how often did it occur at an individual level.
JBI's critical appraisal checklist items | Responses |
---|---|
Is the review question clearly and explicitly stated? | Yes, the authors explicitly stated the research question under review, based on patient, intervention, comparison, outcome and time (PICOT) format |
Were the inclusion criteria appropriate for the review question? | Yes, the inclusion criteria were predefined by the authors and appropriate for the question under review |
Was the search strategy appropriate? | Yes, the search had appropriate terms |
Were the sources and resources used to search for studies adequate? | Yes, authors attempted to identify all the available evidence at the time by searching multiple electronic databases, trial registries and grey literature to minimise publication bias |
Were the criteria for appraising studies appropriate? | Yes, critical appraisal of studies was conducted by two independent reviewers using the Quality Assessment of Diagnostic Accuracy Studies version2 (QUADAS-2) tool |
Was critical appraisal conducted by two or more reviewers independently? | Yes, appraisal of studies was conducted by two independent reviewers using the QUADAS-2 tool and arbitration by a third author if consensus was unable to be achieved by discussion |
Were there methods to minimise errors in data extraction? | No, title and abstract screening were conducted independently by two reviewers. However, the precise methods for full paper review and data extraction remain unclear |
Were the methods used to combine studies appropriate? | No, the methods used by the authors to synthesise the evidence were unclear |
Was the likelihood of publication bias assessed? | No, authors conducted a comprehensive literature search, including attempts to locate grey literature or unpublished studies to minimise selection and publication bias. However, no indication of assessing publication bias of included studies was given |
Were recommendations for policy and/or practice supported by the reported data? | Yes, authors indicated that wearable cuffless blood pressure (BP) devices were promising, however still in their early stages of development as most were prototypes and not available commercially |
Were the specific directives for new research appropriate? | Yes, authors highlighted existing gaps in assessing the validity and usability of cuffless BP devices with lack of standard validation protocols and proposed further research in this domain |
The primary meta-analysis revealed no statistically significant error rates for both systolic and diastolic measurements, with a mean error rate for both systolic and diastolic falling below the mean bias threshold of <5 mmHg. However, there was a large imprecision for systolic blood pressure, with the highest point of the 95% confidence interval being 9.01 mmHg, which could be clinically significant, as previous studies have defined clinically significant differences in blood pressure of error rates of >5 mmHg for systolic blood pressure and >2 mmHg for diastolic blood pressure (Ray et al, 2012).
There was substantial unexplained heterogeneity that suggested important moderating factors that influenced the variation in the mean bias between studies. Previous assessments of cuff-based blood pressure monitoring devices have shown that talking, acute exposure to cold, recent exertion, alcohol consumption and body position can substantially alter the bias within measurements (>5 mmHg) (McAlister and Straus, 2001).
Multiple techniques were combined within the main analysis, which could be a potential cause of the heterogeneity observed between studies. The review conducted a subgroup analysis of the different test types used, where it appears that photoplethysmography alone resulted in a larger mean bias compared to photoplethysmography plus electrocardiogram, for both systolic and diastolic blood pressure. However, because of the unusual method of synthesis for these analyses, the certainty of this difference is questionable.
Regarding external validity of these findings in context to community nursing, it is difficult to establish similarity between the protocols used within the particular studies to those employed during typical clinical practices. The NICE (2023) guidelines recommend taking at least two consecutive measurements, one minute apart twice a day (morning and evening), for blood pressure monitoring at home.
Based on the evidence, the relevant external validity of these findings in context to the current NICE recommendation is unclear. The substantial heterogeneity, imprecision, indirectness and risk of bias observed in the included studies undermine the certainty of the presented estimates to the extent that it is probable that the actual estimates may differ substantially from those presented in this review. Therefore, no clinical recommendations can be made regarding the standard use of cuffless blood pressure devices for community nursing without further research to explore the effects of already known moderating factors.
Conclusions
Future research should focus on comparing multiple cuffless blood pressure devices using photoplethysmography alone and photoplethysmography plus electrocardiogram. It is important that the studies use standardised protocols as suggested by European Society of Hypertension (Stergiou et al, 2023) to allow transparency and repeatability of methods used within the assessment process. Because of the varying methodological issues identified in this review, it is important to repeat the study with greater transparency regarding the inclusion criteria and corresponding methods of synthesis. Relevant moderating factors should be assessed in additional studies to identify the cause of this substantial heterogeneity.
Key points
- Low certainty evidence suggests that wearable cuffless blood pressure monitors may provide an accurate measurement.
- Because of this lack of certainty, no clinical recommendations for community nursing can be made regarding their standard adoption.
- Reliable validation studies are still required, comparing a range of different techniques using a standardised and internationally recognised protocol.
CPD reflective questions
- What moderating factors may affect the accuracy of blood pressure monitoring?
- What methods of remote monitoring are available for your patients?
- Based upon the findings of this review, would you recommend cuffless blood pressure monitoring devices?